Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. An electrical circuit, comprising: a plurality of Retinal Ganglion Cell (RGC) circuits, wherein each of the RGC circuits generates, at an output, a sum of weighted inputs from receptor circuits associated with that RGC circuit; a plurality of primary visual cortex cell (V1) circuits, wherein each of the V1 circuits generates another sum of weighted outputs of a subset of the RGC circuits; and an interface circuit configured to connect two or more of the V1 circuits to each other, wherein the two or more V1 circuits receive inputs from the same RGC circuits of the plurality of RGC circuits, wherein: a weight connecting one of the RGC circuits from the subset with that V1 circuit is positive when modeling that RGC circuit from the subset as an ON RGC, and the weight is negative when modeling that RGC circuit as an OFF RGC.
An electronic circuit mimics the early visual system. It contains multiple Retinal Ganglion Cell (RGC) circuits, each summing weighted inputs from associated receptor circuits. The output of each RGC represents the combined activity of its receptive field. The circuit also includes multiple primary visual cortex (V1) cell circuits. Each V1 circuit sums the weighted outputs from a *subset* of the RGC circuits. Crucially, two or more V1 circuits receive inputs from the *same* RGC circuits. The weights connecting RGCs to a V1 circuit are positive if the RGC is modeled as an "ON" cell (activated by light), and negative if modeled as an "OFF" cell (inhibited by light). This arrangement allows the V1 cells to detect features based on the spatial arrangement of ON and OFF RGC activity.
2. The electrical circuit of claim 1 , wherein the two or more V1 circuits receiving inputs from the same RGC circuits inhibit each other.
The electronic circuit, which mimics the early visual system and contains RGCs and V1 circuits, has this feature: two or more V1 circuits receiving inputs from the same RGC circuits *inhibit* each other. This lateral inhibition between V1 circuits helps to refine feature detection by suppressing weaker or less relevant responses, sharpening the overall response and making the system more selective to specific visual features.
3. The electrical circuit of claim 1 , wherein the two or more V1 circuits receiving inputs from the same RGC circuits inhibit each other in a Winner-Take-All competition.
The electronic circuit, which mimics the early visual system and contains RGCs and V1 circuits, features two or more V1 circuits receiving inputs from the same RGC circuits which inhibit each other via a "Winner-Take-All" mechanism. This means that among the V1 circuits receiving similar input, only the V1 circuit with the strongest response remains active, suppressing the activity of the others. This implements a form of competitive learning and enhances feature selectivity.
4. The electrical circuit of claim 1 , wherein that V1 circuit generates a signal, if the other sum exceeds a threshold.
The electronic circuit, which mimics the early visual system and contains RGCs and V1 circuits, includes a V1 circuit that generates an output signal *only if* the sum of its weighted RGC inputs exceeds a certain threshold. This thresholding mechanism prevents the V1 circuit from responding to weak or irrelevant inputs, reducing noise and enhancing the reliability of feature detection. The V1 circuit effectively acts as a detector firing when sufficient evidence for a particular feature is present in the input.
5. The electrical circuit of claim 1 , further comprising: a circuit configured to adjust weights applied on the outputs for generating the other sum.
The electronic circuit, which mimics the early visual system and contains RGCs and V1 circuits, has an added feature: a circuit that *adjusts* the weights applied to the RGC outputs *before* they are summed by the V1 circuits. This weight adjustment allows the circuit to learn and adapt to different visual patterns by modifying the strength of connections between RGCs and V1 circuits. This enables the circuit to dynamically tune its feature detection capabilities.
6. An electrical circuit, comprising: a plurality of Retinal Ganglion Cell (RGC) circuits, wherein each of the RGC circuits generates, at an output, a sum of weighted inputs from receptor circuits associated with that RGC circuit; a plurality of primary visual cortex cell (V1) circuits, wherein each of the V1 circuits generates another sum of weighted outputs of a subset of the RGC circuits; an interface circuit configured to connect two or more of the V1 circuits to each other, wherein the two or more V1 circuits receive inputs from the same RGC circuits of the plurality of RGC circuits; and a circuit configured to adjust weights applied on the outputs for generating the other sum, wherein the adjustment of each of the weights is based on a sign of one of the outputs associated with that weight.
An electronic circuit mimicking the visual system contains RGC circuits that sum weighted receptor inputs, and V1 circuits that sum weighted RGC outputs, with two or more V1 circuits receiving inputs from the same RGCs. Crucially, a weight adjustment circuit modifies the connection strengths between RGCs and V1 circuits. The *change* in each weight is based on the *sign* (positive or negative) of the V1 circuit's output associated with that weight. This implements a form of Hebbian learning where connection strengths are adjusted based on the correlation between pre- and post-synaptic activity.
7. The electrical circuit of claim 6 , wherein: that weight is increased, if the sign is positive, and that weight is decreased, if the sign is negative.
The electrical circuit described previously adjusts connection weights between RGCs and V1 circuits based on the sign of the V1 circuit output. Specifically, *if the sign is positive, the weight is increased*, strengthening the connection. Conversely, *if the sign is negative, the weight is decreased*, weakening the connection. This reinforcement learning mechanism encourages the V1 circuit to respond more strongly to patterns that elicit a positive output and less strongly to patterns that elicit a negative output.
8. The electrical circuit of claim 1 , wherein that V1 circuit does not generate any signal, if the other sum does not exceed a threshold.
The electronic circuit, which mimics the early visual system and contains RGCs and V1 circuits, includes a V1 circuit that *doesn't* generate any output signal if the sum of its weighted RGC inputs does *not* exceed a specific threshold. This means the V1 cell only responds to sufficiently strong activation patterns, filtering out noise and preventing spurious detections. This implements a minimum activation requirement for feature detection.
9. An electrical circuit, comprising: a plurality of Retinal Ganglion Cell (RGC) circuits, wherein each of the RGC circuits generates, at an output, a sum of weighted inputs from receptor circuits associated with that RGC circuit; a plurality of primary visual cortex cell (V1) circuits, wherein each of the V1 circuits generates another sum of weighted outputs of a subset of the RGC circuits; and an interface circuit configured to connect two or more of the V1 circuits to each other, wherein the two or more V1 circuits receive inputs from the same RGC circuits of the plurality of RGC circuits, wherein the inputs from the receptor circuits are weighted according to a Laplacian window function.
An electronic circuit mimics the early visual system, containing RGC circuits and V1 circuits, where two or more V1 circuits receive inputs from the same RGCs. The circuit weights the inputs from the receptor circuits (before they reach the RGCs) according to a *Laplacian window function*. This function emphasizes changes and edges in the input image by performing a second-order spatial derivative. The RGCs effectively detect local contrast and spatial frequencies.
10. A method for implementing a neural system, comprising: generating, at an output of each Retinal Ganglion Cell (RGC) circuit of a plurality of RGC circuits in the neural system, a sum of weighted inputs from receptor circuits associated with that RGC circuit; generating, by each primary visual cortex cell (V1) circuit of a plurality of V1 circuits in the neural system, another sum of weighted outputs of a subset of the RGC circuits; and connecting two or more of the V1 circuits to each other, wherein the two or more V1 circuits receive inputs from the same RGC circuits of the plurality of RGC circuits, wherein: a weight connecting one of the RGC circuits from the subset with that V1 circuit is positive when modeling that RGC circuit from the subset as an ON RGC, and the weight is negative when modeling that RGC circuit as an OFF RGC.
A method for implementing a neural system mimics the early visual pathway. It involves generating a weighted sum of receptor inputs at the output of each RGC circuit. Then, each V1 circuit generates a weighted sum of outputs from a *subset* of the RGC circuits. Two or more V1 circuits are connected and receive inputs from the *same* RGC circuits. The connections are weighted such that an RGC modeled as "ON" has a positive weight, while an RGC modeled as "OFF" has a negative weight. This allows V1 cells to detect patterns based on the spatial arrangement of light and dark.
11. The method of claim 10 , wherein the two or more V1 circuits receiving inputs from the same RGC circuits inhibit each other.
The method for implementing a neural system which mimics the early visual pathway (using weighted sums in RGC and V1 circuits), includes a step where two or more V1 circuits receiving inputs from the same RGC circuits *inhibit* each other. This lateral inhibition refines feature detection by suppressing weaker responses and enhancing selectivity to specific visual features.
12. The method of claim 10 , wherein the two or more V1 circuits receiving inputs from the same RGC circuits inhibit each other in a Winner-Take-All competition.
The method for implementing a neural system which mimics the early visual pathway (using weighted sums in RGC and V1 circuits) uses a "Winner-Take-All" mechanism for two or more V1 circuits receiving inputs from the same RGC circuits. This means that the V1 circuit with the strongest response suppresses the activity of the others, promoting competitive learning and enhancing feature selectivity.
13. The method of claim 10 , further comprising: generating a signal by that V1 circuit, if the other sum exceeds a threshold.
The method for implementing a neural system which mimics the early visual pathway (using weighted sums in RGC and V1 circuits) includes a step of *generating a signal* by the V1 circuit *only if* the sum of its weighted inputs exceeds a specific threshold. This prevents the V1 circuit from responding to noise and ensures a reliable detection of visual features.
14. The method of claim 10 , further comprising: adjusting weights applied on the outputs for generating the other sum.
The method for implementing a neural system which mimics the early visual pathway (using weighted sums in RGC and V1 circuits) includes a step of *adjusting the weights* applied to the RGC outputs *before* they are summed by the V1 circuits. This allows the system to learn and adapt by modifying connection strengths, dynamically tuning its feature detection.
15. A method for implementing a neural system, comprising: generating, at an output of each Retinal Ganglion Cell (RGC) circuit of a plurality of RGC circuits in the neural system, a sum of weighted inputs from receptor circuits associated with that RGC circuit; generating, by each primary visual cortex cell (V1) circuit of a plurality of V1 circuits in the neural system, another sum of weighted outputs of a subset of the RGC circuits; connecting two or more of the V1 circuits to each other, wherein the two or more V1 circuits receive inputs from the same RGC circuits of the plurality of RGC circuits; and adjusting weights applied on the outputs for generating the other sum, wherein the adjustment of each of the weights is based on a sign of one of the outputs associated with that weight.
A method for implementing a neural system that mimics early visual processing involves: generating weighted sums of receptor inputs at RGC outputs; generating weighted sums of RGC outputs at V1 circuits, with V1 circuits sharing RGC inputs; and *adjusting* the connection weights between RGCs and V1 circuits. Critically, the weight *adjustment* is based on the *sign* (positive or negative) of the V1 circuit's output associated with that weight. This enables learning through correlation between pre- and post-synaptic activity.
16. The method of claim 15 , wherein: that weight is increased, if the sign is positive, and that weight is decreased, if the sign is negative.
The method of implementing a neural system that mimics early visual processing, where connection weights are adjusted based on V1 output sign, works as follows: *if the sign is positive, the weight is increased*, strengthening the connection. Conversely, *if the sign is negative, the weight is decreased*, weakening the connection. This reinforces connections that contribute to a positive output, facilitating learning and adaptation.
17. The method of claim 10 , wherein that V1 circuit does not generate any signal, if the other sum does not exceed a threshold.
The method for implementing a neural system which mimics the early visual pathway (using weighted sums in RGC and V1 circuits) contains a V1 circuit which *doesn't* generate any signal if the sum of its weighted RGC inputs does *not* exceed a threshold. The V1 cell only responds to sufficiently strong activation patterns, filtering out noise and preventing false detections.
18. A method for implementing a neural system, comprising: generating, at an output of each Retinal Ganglion Cell (RGC) circuit of a plurality of RGC circuits in the neural system, a sum of weighted inputs from receptor circuits associated with that RGC circuit; generating, by each primary visual cortex cell (V1) circuit of a plurality of V1 circuits in the neural system, another sum of weighted outputs of a subset of the RGC circuits; and connecting two or more of the V1 circuits to each other, wherein the two or more V1 circuits receive inputs from the same RGC circuits of the plurality of RGC circuits, wherein the inputs from the receptor circuits are weighted according to a Laplacian window function.
A method for implementing a neural system mimics the early visual pathway, containing RGC circuits and V1 circuits, where two or more V1 circuits receive inputs from the same RGCs. The method weights the inputs from the receptor circuits (before they reach the RGCs) according to a *Laplacian window function*. This emphasizes changes and edges in the input image.
19. An apparatus, comprising: means for generating, at an output of each Retinal Ganglion Cell (RGC) circuit of a plurality of RGC circuits in the apparatus, a sum of weighted inputs from receptor circuits associated with that RGC circuit; means for generating, by each primary visual cortex cell (V1) circuit of a plurality of V1 circuits in the apparatus, another sum of weighted outputs of a subset of the RGC circuits; and means for connecting two or more of the V1 circuits to each other, wherein the two or more V1 circuits receive inputs from the same RGC circuits of the plurality of RGC circuits, wherein a weight connecting one of the RGC circuits from the subset with that V1 circuit is positive when modeling that RGC circuit from the subset as an ON RGC, and the weight is negative when modeling that RGC circuit as an OFF RGC.
An apparatus that emulates early visual processing consists of: "means for" generating weighted sums of receptor inputs at RGC outputs; "means for" generating weighted sums of RGC outputs at V1 circuits, where V1 circuits share RGC inputs; and "means for" connecting two or more V1 circuits to each other receiving inputs from the same RGC circuits. RGC to V1 connection weights are positive for "ON" cells and negative for "OFF" cells, enabling feature detection.
20. The apparatus of claim 19 , wherein the two or more V1 circuits receiving inputs from the same RGC circuits inhibit each other.
The apparatus that emulates early visual processing using weighted sums in RGC and V1 circuits, "means for" generating connections, and "means for" weighting includes this: the "means for" connecting two or more V1 circuits receiving inputs from the same RGC circuits causes them to *inhibit* each other, enhancing feature selectivity through lateral inhibition.
21. The apparatus of claim 19 , wherein the two or more V1 circuits receiving inputs from the same RGC circuits inhibit each other in a Winner-Take-All competition.
The apparatus emulating early visual processing with RGCs, V1s, shared RGC inputs, and "means for" summing weighted inputs, incorporates a "Winner-Take-All" competition between two or more V1 circuits receiving inputs from the same RGC circuits. Only the strongest responding V1 circuit remains active, enhancing feature discrimination.
22. The apparatus of claim 19 , further comprising: means for generating a signal by that V1 circuit, if the other sum exceeds a threshold.
The apparatus emulating early visual processing with RGCs, V1s, shared RGC inputs, "means for" summing weighted inputs, further includes "means for" generating a signal by the V1 circuit *only if* the sum of its weighted inputs exceeds a specific threshold. This prevents the circuit from responding to weak or irrelevant inputs, improving the reliability of feature detection.
23. The apparatus of claim 19 , further comprising: means for adjusting weights applied on the outputs for generating the other sum.
The apparatus emulating early visual processing with RGCs, V1s, shared RGC inputs, "means for" summing weighted inputs, further includes "means for" *adjusting the weights* applied to the RGC outputs *before* they are summed by the V1 circuits. This allows the apparatus to learn and adapt by dynamically modifying connection strengths.
24. An apparatus, comprising: means for generating, at an output of each Retinal Ganglion Cell (RGC) circuit of a plurality of RGC circuits in the apparatus, a sum of weighted inputs from receptor circuits associated with that RGC circuit; means for generating, by each primary visual cortex cell (V1) circuit of a plurality of V1 circuits in the apparatus, another sum of weighted outputs of a subset of the RGC circuits; means for connecting two or more of the V1 circuits to each other, wherein the two or more V1 circuits receive inputs from the same RGC circuits of the plurality of RGC circuits; and means for adjusting weights applied on the outputs for generating the other sum, wherein the adjustment of each of the weights is based on a sign of one of the outputs associated with that weight.
An apparatus mimicking the visual system contains "means for" generating weighted sums of receptor inputs at RGCs, "means for" generating weighted sums of RGC outputs at V1s with shared inputs. It also includes "means for" *adjusting weights* between RGCs and V1s, where the *adjustment* is based on the *sign* of the V1 circuit's output associated with that weight, enabling learning based on correlation.
25. The apparatus of claim 24 , wherein: that weight is increased, if the sign is positive, and that weight is decreased, if the sign is negative.
The apparatus described above adjusts connection weights between RGCs and V1 circuits based on the sign of the V1 circuit output. Specifically, the "means for adjusting weights" increases the weight *if the sign is positive* and decreases the weight *if the sign is negative*, facilitating reinforcement learning.
26. The apparatus of claim 19 , wherein that V1 circuit does not generate any signal, if the other sum does not exceed a threshold.
The apparatus emulating early visual processing with RGCs, V1s, shared RGC inputs, and "means for" summing weighted inputs, has the feature that the V1 circuit *does not generate* any signal *if* the sum of its weighted RGC inputs *does not* exceed a threshold. Only strong activation patterns trigger a response, filtering out noise.
27. An apparatus, comprising: means for generating, at an output of each Retinal Ganglion Cell (RGC) circuit of a plurality of RGC circuits in the apparatus, a sum of weighted inputs from receptor circuits associated with that RGC circuit; means for generating, by each primary visual cortex cell (V1) circuit of a plurality of V1 circuits in the apparatus, another sum of weighted outputs of a subset of the RGC circuits; and means for connecting two or more of the V1 circuits to each other, wherein the two or more V1 circuits receive inputs from the same RGC circuits of the plurality of RGC circuits, wherein the inputs from the receptor circuits are weighted according to a Laplacian window function.
The apparatus emulating early visual processing incorporates RGCs, V1s, shared RGC inputs, and "means for" summing weighted inputs. It weights inputs from receptor circuits using a *Laplacian window function* via its "means for" connecting circuits. This emphasizes edges and changes in the input image.
28. A computer program product of a neural system, comprising a non-transitory computer-readable medium comprising code for: generating, at an output of each Retinal Ganglion Cell (RGC) circuit of a plurality of RGC circuits in the neural system, a sum of weighted inputs from receptor circuits associated with that RGC circuit; generating, by each primary visual cortex cell (V1) circuit of a plurality of V1 circuits in the neural system, another sum of weighted outputs of a subset of the RGC circuits; and connecting two or more of the V1 circuits to each other, wherein the two or more V1 circuits receive inputs from the same RGC circuits of the plurality of RGC circuits, wherein a weight connecting one of the RGC circuits from the subset with that V1 circuit is positive when modeling that RGC circuit from the subset as an ON RGC, and the weight is negative when modeling that RGC circuit as an OFF RGC.
A computer program product (stored on a non-transitory medium) for a neural system includes code for: generating weighted sums of receptor inputs at the output of each RGC circuit; generating weighted sums of RGC outputs at each V1 circuit; and connecting two or more V1 circuits to each other, such that they receive inputs from the same RGC circuits. RGC to V1 connection weights are positive for "ON" cells and negative for "OFF" cells, enabling feature detection.
29. The computer program product of claim 28 , wherein the two or more V1 circuits receiving inputs from the same RGC circuits inhibit each other.
The computer program product for a neural system described previously (using weighted sums and ON/OFF coding) includes code such that two or more V1 circuits receiving inputs from the same RGC circuits *inhibit* each other. This implements lateral inhibition, enhancing feature selectivity.
30. The computer program product of claim 28 , wherein the two or more V1 circuits receiving inputs from the same RGC circuits inhibit each other in a Winner-Take-All competition.
The computer program product for a neural system described previously, which uses weighted sums in RGC and V1 circuits, two or more V1 circuits receiving inputs from the same RGC circuits inhibit each other through a "Winner-Take-All" mechanism. This enhances feature discrimination through competition.
31. The computer program product of claim 28 , wherein the computer-readable medium further comprises code for: generating a signal by that V1 circuit, if the other sum exceeds a threshold.
The computer program product for a neural system using weighted sums and ON/OFF coding, includes code for *generating a signal* by the V1 circuit *only if* the sum of its weighted inputs exceeds a threshold. This ensures that the V1 circuit only responds to sufficiently strong activation patterns.
32. The computer program product of claim 28 , wherein the computer-readable medium further comprises code for: adjusting weights applied on the outputs for generating the other sum.
The computer program product for a neural system that implements the visual pathway includes code for *adjusting weights* applied to the RGC outputs *before* they are summed by the V1 circuits. This enables learning and adaptation through dynamic modification of connection strengths.
33. A computer program product of a neural system, comprising a non-transitory computer-readable medium comprising code for: generating, at an output of each Retinal Ganglion Cell (RGC) circuit of a plurality of RGC circuits in the neural system, a sum of weighted inputs from receptor circuits associated with that RGC circuit; generating, by each primary visual cortex cell (V1) circuit of a plurality of V1 circuits in the neural system, another sum of weighted outputs of a subset of the RGC circuits; connecting two or more of the V1 circuits to each other, wherein the two or more V1 circuits receive inputs from the same RGC circuits of the plurality of RGC circuits; and adjusting weights applied on the outputs for generating the other sum, wherein the adjustment of each of the weights is based on a sign of one of the outputs associated with that weight.
A computer program product (on a non-transitory medium) for a neural system mimics early visual processing and contains code for: generating weighted sums of receptor inputs at RGC outputs; generating weighted sums of RGC outputs at V1 circuits, where V1 circuits share RGC inputs; and *adjusting* the weights between RGCs and V1s based on the *sign* of the V1 circuit's output associated with that weight.
34. The computer program product of claim 33 , wherein: that weight is increased, if the sign is positive, and that weight is decreased, if the sign is negative.
The computer program product described above adjusts the connection weights between RGCs and V1 circuits. The program increases the weight *if the sign is positive* and decreases the weight *if the sign is negative*, implementing reinforcement learning.
35. The computer program product of claim 28 , wherein that V1 circuit does not generate any signal, if the other sum does not exceed a threshold.
The computer program product implementing the visual pathway includes code that ensures the V1 circuit *does not* generate a signal *if* the sum of its weighted RGC inputs *does not* exceed a threshold. This prevents spurious responses to weak activation patterns.
36. A computer program product of a neural system, comprising a non-transitory computer-readable medium comprising code for: generating, at an output of each Retinal Ganglion Cell (RGC) circuit of a plurality of RGC circuits in the neural system, a sum of weighted inputs from receptor circuits associated with that RGC circuit; generating, by each primary visual cortex cell (V1) circuit of a plurality of V1 circuits in the neural system, another sum of weighted outputs of a subset of the RGC circuits; and connecting two or more of the V1 circuits to each other, wherein the two or more V1 circuits receive inputs from the same RGC circuits of the plurality of RGC circuits, wherein the inputs from the receptor circuits are weighted according to a Laplacian window function.
The computer program product mimics the early visual pathway and contains code that weights inputs from receptor circuits using a *Laplacian window function*. This enhances edges and changes in the input image through processing in the RGC circuits and ultimately the V1 circuits.
Unknown
September 23, 2014
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.